Do not propogate (pytorch#124769)

Fix the propogate typos.

Pull Request resolved: pytorch#124769
Approved by: https://github.com/Skylion007

test/inductor/test_torchinductor.py

@@ -8612,7 +8612,7 @@ def fn(x):
             self.common(fn, (torch.ones(1, 1, 13, dtype=dtype),))
 
     @unittest.skipIf(not HAS_CPU or not RUN_CPU, "requires C++ compiler")
-    def test_data_type_propogation(self):
+    def test_data_type_propagation(self):
         from torch._dynamo.utils import detect_fake_mode
         from torch._inductor.codegen.common import boolean_ops
         from torch._inductor.compile_fx import _shape_env_from_inputs

torch/_inductor/codegen/common.py

@@ -229,12 +229,12 @@ def deduce_node_dtype_by_inputs(self, node: torch.fx.Node):
         if len(input_nodes) == 0:
             return None
 
-        all_input_nodes_propogated = all(
+        all_input_nodes_propagated = all(
             OptimizationContext.key in n.meta
             and n.meta[OptimizationContext.key].dtype is not None
             for n in input_nodes
         )
-        if not all_input_nodes_propogated:
+        if not all_input_nodes_propagated:
             return None
 
         return functools.reduce(

torch/ao/pruning/_experimental/pruner/prune_functions.py

@@ -84,7 +84,7 @@ def _prune_module_bias(module: nn.Module, mask: Tensor) -> None:
         delattr(module, "_bias")
 
 
-def _propogate_module_bias(module: nn.Module, mask: Tensor) -> Optional[Tensor]:
+def _propagate_module_bias(module: nn.Module, mask: Tensor) -> Optional[Tensor]:
     r"""
     In the case that we need to propagate biases, this function will return the biases we need
     """
@@ -143,7 +143,7 @@ def prune_linear_activation_linear(
     if getattr(linear1, "prune_bias", False):
         _prune_module_bias(linear1, mask)
     else:
-        pruned_biases = _propogate_module_bias(linear1, mask)
+        pruned_biases = _propagate_module_bias(linear1, mask)
         if pruned_biases is not None:
             if activation:
                 pruned_biases = activation(pruned_biases)
@@ -251,7 +251,7 @@ def prune_conv2d_activation_conv2d(
         if prune_bias:
             _prune_module_bias(conv2d_1, mask)
         else:
-            pruned_biases = _propogate_module_bias(conv2d_1, mask)
+            pruned_biases = _propagate_module_bias(conv2d_1, mask)
             if pruned_biases is not None:
                 if activation:
                     pruned_biases = activation(pruned_biases)
@@ -335,7 +335,7 @@ def prune_conv2d_pool_flatten_linear(
     if getattr(conv2d, "prune_bias", False):
         _prune_module_bias(conv2d, mask)
     else:
-        pruned_biases = cast(Tensor, _propogate_module_bias(conv2d, mask))
+        pruned_biases = cast(Tensor, _propagate_module_bias(conv2d, mask))
         flattened_pruned_biases = torch.tensor(
             [[bias] * flatten_scale for bias in pruned_biases], device=mask.device
         ).flatten()

torch/csrc/jit/frontend/ir_emitter.cpp

@@ -5685,7 +5685,7 @@ void runCleanupPasses(std::shared_ptr<Graph>& to_clean) {
   // successive runs of immutable constant prop does not change the graph
   ConstantPropagationImmutableTypes(to_clean);
 
-  // Constant Pooling pass must be after ConstantPropogation, which can create
+  // Constant Pooling pass must be after ConstantPropagation, which can create
   // new constants that needs to be pooled.
   ConstantPooling(to_clean);
 

torch/distributed/_spmd/distribute.py

@@ -350,7 +350,7 @@ def default_factory_op_rule(
 }
 
 
-# Dispatch override for factory ops, as DTensor cannot propogate sharding spec
+# Dispatch override for factory ops, as DTensor cannot propagate sharding spec
 # without DTensor inputs.
 FACTORY_OPS: Dict[torch._ops.OpOverload, Callable] = {
     aten.scalar_tensor.default: default_factory_op_rule,